Orchestrate LLM Skills with Semantic Kernel in Python

1. 1
   Install and Configure Environment

   Install the necessary Python libraries. Semantic Kernel uses an API key for the LLM service (like OpenAI). Create a `.env` file in your project root to store your key securely.
2. 2
   Initialize the Kernel

   The `Kernel` is the core orchestrator in Semantic Kernel. Instantiate it and add a chat service. This connects your application to a specific LLM, like GPT-4, which will be used for reasoning.
3. 3
   Create a Native Plugin

   Plugins are how you extend the AI's capabilities. Create a simple Python class and use the `@kernel_function` decorator on its methods. This exposes your native code to the kernel, making it a tool the AI can use.
4. 4
   Import Plugin and Invoke Kernel

   Add your plugin to the kernel so the AI knows about it. Then, invoke the kernel with a prompt. Semantic Kernel's planner will analyze the prompt, see that it needs to use your `LightPlugin`, execute it, and use the result to form a final answer.

import asyncio
import os
from dotenv import load_dotenv

import semantic_kernel as sk
from semantic_kernel.functions.kernel_function_decorator import kernel_function

# 1. Define a native plugin with a function the AI can call
class LightPlugin:
    """A simple plugin to control a light."""
    is_on = False

    @kernel_function(
        name="change_state",
        description="Changes the state of the light to on or off."
    )
    def change_state(self, new_state: bool) -> str:
        """Sets the light to a new state and returns the status."""
        self.is_on = new_state
        status = "on" if self.is_on else "off"
        # This print statement shows the function was actually executed
        print(f"--- Native function LightPlugin.change_state called: Light is now {status} ---")
        return f"The light has been successfully turned {status}."

async def main():
    # 2. Load environment variables from a .env file
    #    Create a .env file with: OPENAI_API_KEY="sk-..."
    load_dotenv()

    # 3. Initialize the Semantic Kernel
    kernel = sk.Kernel()

    api_key = os.getenv("OPENAI_API_KEY")
    if not api_key:
        print("OpenAI API key not found. Please set it in your .env file.")
        return

    # 4. Add the OpenAI chat service to the kernel
    kernel.add_chat_service(
        "chat_service",
        sk.connectors.ai.open_ai.OpenAIChatCompletion("gpt-4", api_key),
    )

    # 5. Import the plugin into the kernel
    kernel.add_plugin(LightPlugin(), "lights")

    # 6. Run the kernel with a prompt that requires the plugin
    prompt = "Hey, can you turn the light on?"
    print(f"User > {prompt}")

    # The kernel will automatically detect the need for the 'lights.change_state' function
    result = await kernel.invoke_prompt(prompt)

    print(f"Assistant > {result}")

if __name__ == "__main__":
    asyncio.run(main())